Performance of PosteriorPredictiveModel for PredictivePopulationModel is very slow

[DavAug]

Investigate why that is

1. Because random number generator is instantiated multiple times?
2. The way dataframe is created?

Start by comparing perfomance of PredictivePopulationModel and PosteriorPredictiveModel

[DavAug]

It is not the RNG, also the new implementation might speed up things anyway. It seems that the execution time grows exponentially with the number of samples, this might suggest that it's the appending of the rather small chunks of dataframes.

So it may be resolved when we move to xarrays for the sample output.